Technological advances have increased data monitoring in sport, offering potential competitive advantages. Wearable technologies, in particular, generate large volumes of data with untapped potential for athlete monitoring. However, practitioners face challenges in managing and interpreting these data. While commonly used descriptively, there is growing interest in leveraging historical data to inform decisions on training loads and performance. This thesis uses training load data from Global Navigation Satellite Systems (GNSS) and Inertial Measurement Units (IMU) wearable sensors to develop predictive models for Rating of Perceived Exertion (RPE) and Oxygen Uptake (VO2) in team sports. Feature engineering combined with deep learning was employed to enhance prediction accuracy. Results show that integrating domain knowledge with engineered features improves RPE prediction over traditional metrics. A pilot study also demonstrates that GNSS and IMU data can predict breath-by-breath VO2, with linear models performing comparably to deep learning approaches. The findings offer practical tools for coaches and sports scientists, supporting more effective load management and non-invasive, evidence-based performance monitoring in elite team sports.